installation.md

Mesh2SMPL Installation

Prerequisites: 64-bit Windows system, Python, Anaconda

To set up and run the Mesh2SMPL repository using Anaconda, follow the detailed step-by-step instructions below EXACTLY as specified. You can use either PowerShell or the Anaconda Command Prompt to execute these instructions.

Video Instructions: https://youtu.be/elg-wVWFlO0

1. Setup Multiview Images Script

1. Create and activate a conda environment for Python 3.9

   First, create a new conda environment for Python 3.9 and activate it:

   conda create --name multiview python=3.9
   conda activate multiview

2. Clone the repository and navigate into it

   Clone the Mesh2SMPL repository and navigate into the directory:

   git clone --recurse-submodules https://github.com/ddkhen11/Mesh2SMPL
   cd Mesh2SMPL

3. Install dependencies using pip

   Install the necessary dependencies using pip:

   pip install --upgrade setuptools wheel build
   pip install pyembree
   pip install -r requirements-multiview.txt

4. Download the PyOpenGL wheel specific to Python 3.9

   Download the PyOpenGL wheel specific to Python 3.9 from Google Drive. The required file is named PyOpenGL-3.1.7-cp39-cp39-win_amd64.whl. Save this file to a known location on your computer.

5. Install PyOpenGL from the downloaded wheel

   Navigate to the directory where you downloaded the wheel file:

   cd path\to\downloaded\wheel

   Install the PyOpenGL package from the downloaded wheel file:

   pip install PyOpenGL-3.1.7-cp39-cp39-win_amd64.whl

   After that is done, navigate back to the Mesh2SMPL directory:

   cd path\to\Mesh2SMPL

2. Setup Fitting OpenPose to the Multiview Images Script

Due to the complexities of installing and running OpenPose, we will instead use AlphaPose on our multiview images and then convert the AlphaPose keypoints to OpenPose keypoints. The following set of instructions is a detailed guide to installing AlphaPose, so please refer to AlphaPose's GitHub repository if you need to troubleshoot anything.

1. Create and activate a conda environment for Python 3.7

   If your previous conda environment is still active, make sure you deactivate it first:

   conda deactivate

   Then create a new conda environment for Python 3.7 and activate it:

   conda create --name alphapose python=3.7
   conda activate alphapose

2. Install dependencies using pip

   Navigate back to the Mesh2SMPL directory and install the necessary dependencies using pip:

   cd path\to\Mesh2SMPL
   pip install -r requirements-pose.txt

3. Delete necessary lines and files from AlphaPose directory

   In third_party/AlphaPose open setup.py and delete line 211. Don't forget to save the file before exiting.

   In the same folder, delete the setup.cfg file.

4. Build and install AlphaPose

   Navigate to the AlphaPose directory, build and install AlphaPose, and navigate back to the Mesh2SMPL directory. Ensure that Microsoft Visual C++ 14.0 or greater has been installed through Microsoft C++ Build Tools https://visualstudio.microsoft.com/visual-cpp-build-tools/ prior to running this:

   cd third_party/AlphaPose
   python setup.py build develop
   cd ../..

5. Download pretrained models

   First, download the YOLO object detection model from this link and place the file in third_party/AlphaPose/detector/yolo/data.

   Then, download the pretrained Fast Pose pose estimation model from this link and place the file in the third_party/AlphaPose/pretrained_models directory.

3. Setup rendering the SMPL Model

1. Create and activate a conda environment for Python 3.6

   If your previous conda environment is still active, make sure you deactivate it first:

   conda deactivate

   Then create a new conda environment for Python 3.6 and activate it:

   conda create --name fit-smpl python=3.6
   conda activate fit-smpl

2. Install dependencies

   Install the necessary dependencies:

   python -m pip install --upgrade pip
   pip install -r requirements-smpl.txt
   conda install pytorch-cpu==1.0.0 torchvision-cpu==0.2.1 cpuonly -c pytorch

3. Download and clean the SMPL model files

   Go to https://smpl.is.tue.mpg.de/, make an account, and download version 1.0.0 of SMPL for Python Users from the downloads page. Extract this zip file. In the extracted folder, go to SMPL_python_v.1.0.0/smpl/models. Copy and paste the files in this folder to tools/smpl_models in your Mesh2SMPL directory.

   Then, go to https://smplify.is.tue.mpg.de/, make an account, and download SMPLIFY_CODE_V2.ZIP from the downloads page. Extract this zip file. In the extracted folder, go to smplify_public/code/models. Copy and paste basicModel_neutral_lbs_10_207_0_v1.0.0.pkl to tools/smpl_models in your Mesh2SMPL directory.

   Finally, run the following command (make sure your current directory is still Mesh2SMPL):

   python tools/clean_models.py --input-models tools/smpl_models/*.pkl --output-folder third_party/MultiviewSMPLifyX/smplx/models/smpl

4. Download pretrained VPoser models

   Go to https://smpl-x.is.tue.mpg.de/, make an account, and download VPoser v1.0 - CVPR'19 (2.5 MB) from the downloads page. Extract this zip file. In the extracted folder, go to vposer_v1_0/snapshots. Copy and paste the *.pt files to third_party/MultiviewSMPLifyX/vposer/models/snapshots in your Mesh2SMPL directory.

After all of this setup, you are ready to run Mesh2SMPL. Follow the instructions in run.md.